Re: AIX 4.3 getaddrinfo busted

From: Bruce Momjian <bruce(at)momjian(dot)us>
To: Andrew Chernow <ac(at)esilo(dot)com>
Cc: PostgreSQL-development <pgsql-hackers(at)postgresql(dot)org>
Subject: Re: AIX 4.3 getaddrinfo busted
Date: 2009-01-23 14:41:56
Message-ID: 200901231441.n0NEfuA22244@momjian.us
Views: Raw Message | Whole Thread | Download mbox | Resend email
Thread:
Lists: pgsql-hackers

Andrew Chernow wrote:
> Bruce Momjian wrote:
> > Andrew Chernow wrote:
> >> AIX 4.3 was released in late 1999, so I thought it was worth mentioning.
> >> I only have AIX 4.3 and 6.1, so I have no idea how AIX 5 handles this.
> >> AIX 6.1 works fine.
> >>
> >> Anyways, the service argument to getaddrinfo is busted on AIX 4.3, thus
> >> src/backend/libpq/ip.c pg_getaddrinfo_all() is busted on this
> >> platform. It fails with EAI_NODATA "Host not found". If this argument
> >> is left NULL, everything works.
> >>
> >> I can supply a patch to fix this. My suggestion would be to always
> >> supply a NULL service to getaddrinfo. After a successful call, set the
> >> port if it was provided ... htons((unsigned short)atoi(servname)). This
> >> approach avoids a configure check.
> >
> > Why would we risk breaking other platforms to avoid an AIX bug? At best
> > we can put a code comment in that section of the code.
> >
>
> IMO, there is no risk. getaddrinfo allows a NULL second argument on
> every platform I have worked with. I can't see this breaking anything.
> Our internal libraries do it this way for this exact reason, and it
> works on a large number of platforms.
>
> But, it can be done so that it only affects AIX machines. There is
> already a #ifdef _AIX block in this function so we can handle all AIX
> versions as I have suggested and let everyone else run the same old code.
>
> #ifdef _AIX
> getaddrinfo(hostname, NULL /* servname */, ....);
> // manually set the port
> #else
> getaddrinfo(hostname, servname, ....); /* same code as now */
> #endif

Well, the platform is from 1999 and we have never had bug reports on
this, even though the code has been there for quite a few years; I just
don't see the motivation for the change.

If you really want this platform to work, I would submit a patch that
tests for a C compiler symbol or #define that is only defined for that
platform and make service = null in that case.

--
Bruce Momjian <bruce(at)momjian(dot)us> http://momjian.us
EnterpriseDB http://enterprisedb.com

+ If your life is a hard drive, Christ can be your backup. +

In response to

Responses

Browse pgsql-hackers by date

  From Date Subject
Next Message Bruce Momjian 2009-01-23 14:43:21 Re: AIX 4.3 getaddrinfo busted
Previous Message Andrew Chernow 2009-01-23 14:32:54 Re: AIX 4.3 getaddrinfo busted